Namespace prefixes are a security risk

["Costello, Roger L." <costello@mitre.org>]

Hi Folks,


INTRODUCTION

The problem described below occurs with XML 'guards' that are trying to prevent the release of unauthorized information at an enclave boundary. Namespace prefixes provide a ready channel for transmitting information out of the protected enclave.  That channel is overlooked by most XML applications, expect for an application that is specifically looking for that information.


THE PROBLEM

Consider this XML document containing data about a book:

    <book xmlns="http://www.book.org";>
        <title>The Origin of Wealth</title>
        <author>Eric D. Beinhocker</author>
        <date>2006</date>
        <ISBN>1-57851-777-X</ISBN>
        <publisher>Harvard Business School Press</publisher>
        <cost currency="USD">29.95</cost>
    </book>

Seems pretty innocuous, right?

It uses a default namespace declaration. Alternatively (equivalently) each element can be qualified. And the prefix can be anything, e.g.,

    <attackNOW:book xmlns:attackNOW="http://www.book.org";>
        <attackNOW:title>The Origin of Wealth</attackNOW:title>
        <attackNOW:author>Eric D. Beinhocker</attackNOW:author>
        <attackNOW:date>2006</attackNOW:date>
        <attackNOW:ISBN>1-57851-777-X</attackNOW:ISBN>
        <attackNOW:publisher>Harvard Business School Press</attackNOW:publisher>
        <attackNOW:cost currency="USD">29.95</cost>
    </attackNOW:book>

Not so innocent-looking anymore, is it?

But the problem isn't in how it "looks." The problem is that, as far as XML tools are concerned, the two forms are exactly equivalent:

- If the first form is schema-valid, 
  then the second form is schema-valid. 

- If the first form can be parsed by an 
  XML parser, then the second form can 
  be parsed by an XML parser.
 
- If the first form can be processed by 
  an XSLT transform, then the second form 
  can be processed by an XML transform.

So, a guard, processing XML documents, using XML tools, may be completely oblivious to the covert information being passed via the namespace prefix. You may even say that the prefix is "invisible" to the guard.


REPLACE NAMESPACE PREFIXES

The good news is that, whatever prefixes the XML document contains, they can be replaced with controlled prefixes. Here is an XSLT transform that replaces the namespace prefix with N103: (Thanks to Ken Holman for this XSLT)

<?xml version="1.0" encoding="US-ASCII"?>
<xsl:stylesheet xmlns:xsl="http://www.w3.org/1999/XSL/Transform";
                version="1.0">

    <xsl:param name="use-this-prefix"/>

    <xsl:template match="*[namespace-uri(.)]">
        <xsl:element name="{$use-this-prefix}{local-name()}"
                     namespace="{namespace-uri(.)}">
            <xsl:apply-templates select="@*|node()"/>
       </xsl:element>
    </xsl:template>

    <xsl:template match="@*[namespace-uri(.)]">
       <xsl:attribute name="{$use-this-prefix}{local-name()}"
                      namespace="{namespace-uri(.)}">
           <xsl:value-of select="."/>
       </xsl:attribute>
    </xsl:template>

    <xsl:template match="@*|node()"><!--identity for all other nodes-->
       <xsl:copy>
           <xsl:apply-templates select="@*|node()"/>
       </xsl:copy>
    </xsl:template>

</xsl:stylesheet>


COMMENTS

I welcome your comments.

/Roger